HI,
> Pei Wang's uncertain logic is **not** probabilistic, though it uses
> frequency calculations
IMO Pei's logic has some strong points, especially that it unifies
fuzzy and probabilistic truth values into one pair of values. I
think in Pei's logic the frequency f is indeed a probability. In
your logic, do you treat fuzziness and probabilities separately?
Like NARS, PLN encompasses both fuzziness and uncertainty, using
different but interoperating inference rules
It seems that a term cannot be uncertain. For example, "women have
long hair" may be associated with a probability, but "women" as a
term does not seem to be probabilistic. I wonder why you say terms
can have truth values?
In a given sample space S, we may say e.g.
P(woman) = .51
P(man) = .49
This sort of expression occurs all the time in probability theory
textbooks.
However, you can rewrite these as
P(woman|S) = .51
P(man|S) = .49
But, NARS does not handle things this way. So, for instance, if you
know the truth value
cat ---> animal
and want to estimate the truth value of
animal --> cat
then PLN in Novamente uses Bayes rule which relies on
P(animal | cat)
P(animal)
P(cat)
to estimate
P(cat | animal)
but NARS, used in the simplest and most direct way, just uses [what
might be written]
inheritance(animal | cat)
to estimate
inheritance(cat | animal)
However, through longer chains of inference, NARS may potentially
draw information about
the frequency of cats and animals in various contexts into its
estimate of the truth value of the
latter relationship. This is an issue Pei and I have discussed
frequently without reaching any
agreement.
My question is: If I provide a set of facts/rules associated with
Pei-style uncertain values (2 numbers), would Novamente find it
useful? Is it easy to convert that into NM's PLN?
Yes. NM and PLN support multiple truth value representations,
including the NARS-like representation, single-number probabilities,
Walley's imprecise probabilities, and also a novel format called
indefinite probabilities. Facilities for interconversion are provided.
-- Ben
-----
This list is sponsored by AGIRI: http://www.agiri.org/email
To unsubscribe or change your options, please go to:
http://v2.listbox.com/member/?list_id=303
